Adds better UX for the user.

Moved from DREAM3D Issue 651

Selecting a DataArrayPath in the GUI does not gray out DataContainers without valid AttributeMatrices.

This allows easier and intuitive selection of data arrays, attribute matrix, data container and geometry .

Composite prebuilt pipeline (Synthetic #5) crashes on launch with an error dialog that reads:

"Internal error gathering statistics from Statistics Widgets. Error code -1000"

This is on Mac OS X Yosemite.

Tested on Windows; it seems that the SIMPLib.dll is not getting into the zipped package. Most recent nightly build reflects this issue.

Something strange is happening with the "Delete Data" filter. I keep getting the error that the arrays I have selected to be deleted ".. cannot be removed because it was not found in CellFeatureData"! See the attached json file (I changed its extension into txt to trick the filter).
ANG (SqrGrid) To Statistics.txt

The output widget resizes larger than the input widget

The current deepCopy() implementations for IGeometry objects do not actually perform a deep copy; instead, they just increase the reference count for each of the underlying arrays.

The only reason we have boost in SIMPL is the Random number generators but this is now all available in the std::namespace by using C++11

ExtractComponentAsArray filter does not throw errors if a negative component ID is specified and fills the initial values in the new array with preexisting data depending on the ID. Larger negative numbers offset the extracted array by larger amounts.

The Qt NTFS file permissions checking is notoriously slow on Windows (see http://doc.qt.io/qt-5/qfiledevice.html for more information). It also seems to sometimes make mistakes with whether the user actually has permissions, preventing writing to certain drive setups. Although it is mostly safe, adding a CMake variable that could turn off the permissions checking would give developers the option to disable the checking for more speed (this is particularly apparent whenever a preflight occurs with a DataContainerWriter filter in the pipeline.

Phase Type Selection Widget not deserializing from pipeline file correctly.

If a filter gets dragged to the bottom of the pipeline widget on Windows, it may get stuck at the bottom of the pipeline widget where the pipeline will be built up from the bottom instead of the top.

char/unsigned char are undefined behavior for std::uniform_int_distribution; instead allow the class to perform type deduction on the constructor ranges and implicitly cast the return type.

The error will be cleared if the DataArray in question exists, but the red background color remains.

Alternative way to set an object into a "*Selection" widget

Slack.org messaging protocol is very easy to use. Create a filter that would allow a user to post a message to their own slack channel.

WriteASCIIData can export StringArrays but ReadASCIIData cannot read them.

DataArray objects currently accept a forceNoAllocate argument on deepCopy(), which gives the option to avoid actually performing an allocation; this is useful if only a "shallow" version of the array is needed, i.e. when only the descriptive information of the array is desired for something like preflight. DataContainers, AttributeMatrices, and IGeometry objects do not have this option, meaning that if one calls deepCopy() on a DataContainer, all DataArrays in that DataContainer end up being copied and allocated. For consistency, add similar "forceNoAllocate" arguments to all the different data structure objects to allow "shallow" copies (the argument can take a default false, just like for DataArray).

Start a DREAM.3D session with a bookmark that is missing. Then on the file system put the bookmark back in the proper spot. In the Bookmarks Widget the row is RED and should update to normal color but does not.

It will work if the bookmark is available on start and then the user deletes the bookmark manually through Windows Explorer or macOS Finder.

Actual C++ name is "RemoveArrays". If someone wants to delete 60 data containers then they have to click 60 different buttons. Need a "select all" checkbox like on other widgets.

The idea of the widget is to give users a more direct connection between the various widgets that allow the user to select input/output data and the actual data hierarchy. This functionality would hopefully lessen some of the difficulties that users have when trying to understand what to select in those input widgets.

Generic functions for computing the "size" of elements for a given geometry should be added for all IGeometry classes. This addition would make filters that compute feature/ensemble size information function on any geometry, not just images. The "size" value will be relevant to the geometry topology (e.g., lengths for the edge geometry, areas for the triangle geometry, volumes for the rectilinear grid geometry, etc.).

Previously, when the pipeline was saved into the output HDF5 .dream3d file, the entire processing history for the data was "saved", since any future processing that used that .dream3d would just append those filters to HDF5 group.

The new way of storing the pipeline just stores the JSON information as one long string. While more compact, this string is overwritten whenever the .dream3d file is used in another pipeline. This is a critical problem, since all processing history for a given .dream3d is lost.

A possible solution is to create a new Pipeline HDF5 data string set within the Pipeline group whenever the .dream3d file is used in a pipeline. This would also require naming the Pipeline string data sets with some numerical counter ("Pipeline_0, Pipeline_1" etc.). Additionally, the functionality used to extract the saved pipelines out of the .dream3d file would have to allow for the option to choose which pipeline to extract, or to extract all of them appended to one another (which was the former behavior).

The AttributeMatrixProxy class does not export its symbols, causing a linker error on Windows platforms.

To recreate: add exactly one filter to a pipeline, click and hold to drag the filter, the drop it in some other location in the pipeline. This elicits a crash in SVPipelineViewWidget (I think on line 680?). The issue is most likely related to the fact that only one filter is in the pipeline when the drag/drop is taking place, so something is probably null.

QtSStringEdit should remove Apply/Cancel buttons and return text color to black when the text field is equal to the last saved value. Currently, this only happens when Apply or Cancel buttons are clicked.

Currently, all these objects implement a deepCopy() function that (should) explicitly copy all contents into a new shared pointer instance (the IGeometry versions are bugged at the moment). The IDataArray implementation allows the developer to avoid an allocation with a Boolean argument (a "shallow" copy), but the other objects do not. This is useful when the descriptive information of the object is desired, but the developer wants to avoid allocating until later.

For clarity, implement shallowCopy() functions for all these objects and remove the deepCopy() argument in IDataArray.

After running a pipeline, the Data Browser is cleared when changing filters until the next preflight is run.

The user can select multiple filters but they can not drag them within the same pipeline.

Changing the DataArrayPath on a DataArray accessed by WriteASCIIData causes a crash. This can most easily be performed by creating the DataArray with a different name or deleting the filter that creates the Data Container / Attribute Matrix / Data Array.

This causes a null pointer in dataCheck() that is never handled.

The current implementation of RayIntersectsPlane takes the plane as defined by the coordinates of three points. Sometimes it is more convenient to have the plane defined by the parameters of the plane equation (point-normal form); the actual algorithm is fundamentally similar. An overload for RayIntersectsPlane should be added that takes this formulation for a plane.

Both ScalarTypeWidget and NumericTypeWidget make use of a C++11 strongly typed enum for their internal parameter representation. In the filterNeedsInputParameters() function, these widgets pull the index of the user selection from the GUI as an int then use the default constructor to construct a QVariant. This QVariant is then set as a property on the filter; the property is stored as an instance of the strongly typed enum.

In C++11, these scoped enums do not decay to integral types automatically, so QVariant(int i) is not equivalent to QVariant(SIMPL::ScalarTypes::Type i). Additionally, SIMPL::ScalarTypes::Type is a user declared QMetaType, and so creating a QVariant for it should use the from/set value approach.

To fix this, replace the QVariant v(i) line in ScalarTypeWidget's and NumericTypeWidget's filterNeedsInputParameters() function with the following:

SIMPL::ScalarTypes::Type type = static_cast<SIMPL::ScalarTypes::Type>(i);
QVariant v;
v.setValue(type);

Prebuilt Pipelines and examples are saved in preferences, which stores an absolute path to the pipelines. Moving the application, leaves the paths to search in the original location of the application rather than using the directory relative to the new application path.

The ImageGeom getCoords functions return plane-centered coordinates, while the RectGridGeom getCoords functions return cell-centered coordiates. The ImageGeom getCoords functions should be updated to properly return cell-centered coordinates. Additionally, a new set of coordinate functions that return plane-centered (i.e., actual nodal coordinates for each voxel) coordinates should be added to the IGeometryGrid interface.

I try to read the DREAM3D file (select file in "Read DREAM..." filter), although I use "Overwrite Existing ...", it cannot be read.

A LinkedDataContainerSelectionFilterParameter/Widget would show/hide other filter parameter widgets based on the geometry type of the selected DataContainer, analogous to how the LinkedChoicesFilterParameter currently works.

The current filter parameter new functions expect an integer for the "group index" used to switch hiding/showing. This naturally works well for the LinkedChoicesFilterParameter; without changing any code, this same logic can be used by static casting the geometry type enum desired to an integer and using that value. The widget then just needs to parse the current geometry selection and emit the integer version of the geometry type enum. This avoids refactoring any filter parameter new functions, but is somewhat messy.

More generically, it may be beneficial in the future to have all object selection widgets (array selection, etc.) be capable of linking to hide/show other widgets based on the objects requirement type (primitive type, component dimensions, etc.). The data container version of this is the simplest, since only the geometry type can be composed into the requirement type.

Show the selection/filtering criteria for a specific data selection widget, i.e., why are only certain arrays, attribute matrix or data containers allowed. This will need some interesting diagnostics to be used so we can tell the user why a selection failed.

User had a use case where we needed to move 8 data arrays from one attribute matrix to another attribute matrix. 8 instances of the filter are needed.

Would be nice to have a way to select multiple arrays to move, or multiple of anything to move.

The function MatrixMath::Invert3x3 behaves incorrectly as it always returns transposition of the correct result. For example, matrix (1,0,1,0,1,0,0,0,1) should be inverted as (1,0,-1,0,1,0,0,0,1), but the function returns (1,0,0,0,1,0,-1,0,1) (regardless whether the numbers are placed by rows or by columns).

The current scoped enum for IGeometry types maps the geometry to an unsigned int that is representative of the Vtk cell type; this means that Image and RectGrid share the same enum value (11, for the voxel Vtk cell type). This is a problem when expecting the types to be unique, as would be expected of such an enum declaration.

This can be addressed by renaming the current type enum to IGeometry::VtkCellType, and making a new IGeometry::Type enum with distinct values.

The findDerivatives algorithm is correctly coded for all the different 2D cases in ImageGeom and RectGridGeom (and executes properly in serial) but will crash when the structure is 2D in the XY plane when tbb attempts to partition jobs to threads. The exception occurs because the grain size for the z step is set to the z dimension divided by the default number of threads to obtain optimal striding, but this could lead to a grain size of zero if the number of threads is more than 2.